Related Experiment Videos
Multimodal Deep Learning Approaches for Lung Disease Detection: A Review
Bastian Estay Zamorano1, Ali Dehghan Firoozabadi1, Pablo Adasme2
1Department of Electricity, Universidad Tecnológica Metropolitana, Santiago 7800002, Chile.
Medicina (Kaunas, Lithuania)
|July 28, 2026
Summary
This review synthesizes multimodal deep learning (DL) for lung disease diagnosis using imaging, acoustics, and electronic health records (EHRs). It highlights dominant DL models and identifies barriers to clinical translation.
Area of Science:
- Pulmonary Medicine
- Artificial Intelligence
- Medical Imaging
Background:
- Lung diseases are a major global health burden.
- Current deep learning (DL) reviews for pulmonary diagnosis often lack multimodal integration.
- Integrating imaging, acoustic, and electronic health record (EHR) data offers a comprehensive approach to lung disease detection.
Purpose of the Study:
- To review the state-of-the-art in multimodal deep learning for lung disease detection and classification.
- To identify dominant DL architectures, performance benchmarks, and translational challenges.
- To cover advancements from 2019 to 2024 across various data modalities.
Main Methods:
- A structured narrative review of peer-reviewed studies was conducted.
- Searches were performed in PubMed, Scopus, IEEE Xplore, and Web of Science.
- Data extraction focused on performance metrics, dataset characteristics, and limitations.
Main Results:
- Convolutional Neural Networks (CNNs) and Transformers show high performance in chest X-ray classification.
- Acoustic analysis using spectrograms and self-supervised learning (e.g., Wav2Vec 2.0) demonstrates promising, yet dataset-dependent, results.
- Multimodal DL frameworks are emerging but face translational barriers.
Conclusions:
- Multimodal deep learning shows significant potential for improving lung disease diagnosis.
- Further research is needed to overcome dataset dependency and clinical implementation challenges.
- Integrating diverse data sources enhances the accuracy and scope of pulmonary diagnostic AI.